Behavioral science initiatives to reduce individuals' climate footprint

We will differentially expose participants to messaging about their consumption through a mobile app.

2024-05-13

2025-02-01

Consumption
Political behavior and attitudes

Consumption will be measured using bank transaction data and register data. We will measure consumption in terms of carbon footprint and expenditures/purchases in various categories.
Political behavior and attitudes will be measured using register and survey data. We will link register data on voting participation as a binary measure of political behavior. We will also provide subjects an opportunity to donate to an environmental cause and to sign an environmental petition as measures of political behavior. In terms of the survey, we will ask participants about their political preferences, their voting choice in the EU parliament elections, their opinion on environmental policies, and their opinion on the role of government and individual responsibility for confronting societal challenges. As an additional measure of political behavior, we are planning follow-ups with subjects to learn whether they have taken part in a climate demonstration.

Green investments
Carbon literacy
Perceived behavioral plasticity and intention to reduce carbon footprint

Investments (e.g., stock portfolio, pension fund choices) will be observed through register data. We will construct a measure of how green the participant's investments are.
After exposure to treatment, we will gather survey data from participants about their knowledge of the carbon intensiveness of common activities. This outcome will help shed light on users’ learning and the informational value of the treatment.
As part of the treatment, users will be presented with a set of “action cards” that describe measures they can take to reduce their carbon footprints. We will observe and analyze their choice of which action cards they choose to save, which indicates the actions they would like to undertake, in whole or in part, in the next 6 months. We can subset our analysis of consumption outcomes based on the action cards saved by users, which may shed light on mechanisms for the treatment effect.

We will differentially expose participants to messaging about their consumption through a mobile app. Participants will be randomized into two versions of the mobile app: (i) a carbon footprint calculator and (ii) a financial management calculator. Both versions of the app will have a similar look and feel, but the treatment (carbon footprint calculator) will expose participants to personal responsibility messaging for addressing climate change.

We will recruit participants from the Swedish population to join our study and download the app "Konsumtionskollen" ("Consumption checker"). We will conduct an initial baseline survey on users’ attitudes and behaviors about the environment and other political and social issues. We will also randomize users into two different versions of the app. Before treatment assignment, users will fill out a number of survey questions to provide some baseline information. Some of the pre-treatment information that we gather will be used for heterogeneity analysis: likely green voters, desire who to live in an environmentally friendly way, participation in past protests, trolley dilemma responses, pre-treatment carbon footprint, whether the user underestimated their own carbon footprint, environmental values, environmental self-identity, and the degree to which the user thinks climate change is a serious problem.
The treatment group's version is an adaptation of Svalna, a carbon footprint app that provides feedback on the user's consumption. The app links with users’ BankIDs (with informed consent) and leverages transaction data to show users their carbon footprint based on their purchasing habits. The user can get tips on how to reduce their carbon footprint, set their own reduction targets, and compare their emissions with peers. The app also presents users with “action cards” that describe measures they can take to reduce their carbon footprints. Users can save these action cards to indicate the actions they would like to undertake, in whole or in part, in the next 6 months.
The control group will use a version of the app that is as similar as possible to the Svalna app, but without environmental messaging. Instead, we provide the user feedback on consumption in a similar format, but we focus the messaging on personal financial management. By removing information about the environmental consequences of purchases, we have a clean control group that helps us isolate the effect of personal environmental responsibility messaging on subsequent attitudes and behavior.
We will have two recruitment efforts. One recruitment effort will seek a representative sample of the Swedish working age population living in single households. The other recruitment effort will be targeted toward demographics and geographies that we believe will be most likely to respond to treatment. Note that treatment will be randomized within each recruitment effort, so that we have internal validity for both.

randomization done in office by a computer

individual

No

4500 individuals

4500 individuals

2250 individuals control app; 2250 individuals treatment app

With a sample size of 4000 individuals and 80% power, we will have an MDE for consumption (carbon footprint) on the order of 0.04 log points. In pilot data, the mean for ln(carbon footprint + 1) is 4.14 and the standard deviation for this outcome is 1.37 log points, so our MDE is roughly 3% of a standard deviation.